Like the ‘Sanskrit’ and ‘Persian Cosmopolis’, can AI become a transregional force without remaining tied to technological giants of the United States? This article explores the possibilities of the emergence of a vernacular AI cosmopolis, similar to vernacular literature of 15-16 centuries that is locally rooted, multilingual and plural.
Vipul Singh
Introduction
Every age develops its own technologies of knowledge. In premodern India, Sanskrit was a dominant medium of literary, intellectual and political expression. It travelled across regions and connected courts, scholars and intellectual traditions. Sheldon Pollock called this formation the ‘Sanskrit cosmopolis’. Later, Persian acquired a similarly remarkable trans-regional reach. It became a language of administration, literature and political culture across a large part of Asia. In India too, Persian became a language of admiration after the Turks established their rule in Delhi. Pollock used the term ‘Persian cosmopolis’ to describe this wider phenomenon. But very soon after the Taimur invasion and disintegration of Delhi Sultanate rule, various vernacular literature were composed at regional courts, what has been termed ‘vernacular millennium’.
In early India, access to Sanskrit learning was certainly socially and institutionally structured. It depended on educational traditions, teacher-student lineages, institutions, patronage and social position. Sanskrit was known to only those who had access to the institutions, training and intellectual authority associated with it.
Today, we may be witnessing the emergence of another kind of cosmopolitan knowledge system in the form of AI (artificial intelligence). The comparison may sound strange. Sanskrit, Persian and vernacular were languages, whereas AI is a technology. While the languages travels through manuscripts, teachers, courts, saints and intellectual lineages, AI travels through algorithms, data centres, networks and enormous computing power.
Yet there is a deeper connection. All of them are ways of organising and circulating knowledge. And all of them raise similar question. Who gets access to powerful knowledge? And who gets to command it? Lets break this down.
Has Knowledge been equally accessible?
The internet promised to democratise knowledge. A student in a small town could listen to lectures from the world’s best universities. A researcher could access digital libraries from almost anywhere. Someone who could not afford private tuition could learn online. Generative AI has taken this much further. A student can ask AI to explain a difficult concept. A researcher can ask it to compare arguments. Someone who does not speak English well can translate a difficult text. A teacher can generate exercises and examples. A small business can get help with coding, marketing or accounting. Thus, for the first time, sophisticated intellectual assistance can potentially sit inside the pocket of an ordinary person. This is revolutionary.
But having access to AI is not the same as having access to the full power of AI. Having AI is not the same as commanding AI. Imagine two students with the same smartphone and the same free AI service. One asks simple questions and accepts whatever answer appears. The other knows how to ask better questions with advanced versions because she can afford to pay for the subscriptions and combines several AI tools. So technically, both have AI. But they do not possess the same AI.
There is another form of divide between people who have AI and those who do not have it at all. UNESCO too has warned about an emerging AI divide. Nearly one-third of the world’s population still lack internet access. The inequalities in connectivity, skills and education also means that there is a huge population who is bereft of access to AI and its benefits.
Another development that makes this question even more important is that AI itself is becoming stratified. Basic AI services may remain free or relatively inexpensive. But the most powerful models, higher usage limits, advanced reasoning, research capabilities and autonomous tools are increasingly placed behind premium subscriptions. Whether it is Anthropic’s Claude, ChatGPT or Google’s AI Gemini. Their premium versions offer unlimited usage and access to their most capable models and features.
This does not mean that poorer people will have no AI. The situation is more subtle. There may be AI for everyone, but not the same AI for everyone.
From Sanskrit cosmopolis to AI cosmopolis
This is where Pollock’s historical vocabulary becomes useful. The Sanskrit cosmopolis was not simply a collection of people who happened to know Sanskrit. It was a transregional cultural formation. Sanskrit carried literary and political authority across large geographical distances. It created a shared intellectual and cultural space. Persian later acquired a similarly impressive transregional reach.
AI is beginning to acquire some of these characteristics. The most powerful AI models are not tied to one locality. They operate across borders, disciplines and increasingly across languages. They can become common infrastructures for research, education, business and creativity. We can therefore imagine the emergence of an AI cosmopolis. It could be global in reach. But it may not be equal in access. In other words, AI is progressively becoming cosmopolitan in reach while remaining unequal in practice.
A recent India Today opinion piece provides an interesting illustration of another kind of technological hierarchy. Writing about Apple’s latest iPhone pricing in India, the author of the piece describes Apple, tongue-in-cheek, as ‘Manuvadi Apple’ and its pricing strategy as a kind of ‘cost census’. The satire is that the iPhone gradually lost some of its exclusivity as ownership spread across social classes. But Apple’s new pricing restores some of that exclusivity by making the newest models increasingly expensive.
The iPhone metaphor points to something real. Technology can become a marker of social status because access to it is unequal. With the iPhone, the hierarchy is primarily about consumption. With AI, the stakes could be much higher. The hierarchy could concern knowledge production itself. Every day new high-tech devices and cars are being launched and they pose the same question – how much technology can one afford to own? AI is increasingly becoming the most powerful technology and thus asks how much intelligence can one afford to access. As a history practitioner, to me it is a much more consequential question.
Danger of a new knowledge elite
Imagine a future in which AI becomes extremely expensive. Large corporations, elite universities and wealthy individuals would have access to the most powerful systems. Governments would use them for research and policy. Well-funded institutions would employ AI agents capable of conducting complex analysis around the clock. But ordinary citizens would have access only to basic AI models.
The inequality AI may create could become a cognitive infrastructure divide. Those at the top would possess tools that amplify their ability to learn, research, create and solve problems. Their existing advantages could therefore become even larger. In other words, better education, better use of AI, greater productivity, greater income, access to better AI, and still greater productivity. The danger is, therefore, not simply that some people will own better gadgets. Consequences are going to be much graver. AI technology could amplify existing inequalities because of its capacity to produce knowledge.
Could AI be vernacularised?
There was a major transformation in South Asian literary culture from around the second millennium. Vernacular languages increasingly became important vehicles of literary and political expression. Francesca Orsini has very persuasively explained how South Asian literary cultures became multilingual, multifunctional and multilocational. Different languages could coexist. They could serve different purposes. Literary production could happen in courts, towns, religious centres and other social spaces. Texts could circulate through manuscripts, performance and oral traditions. A vernacular did not simply become a local replacement for Sanskrit. Different languages and literary cultures interacted with one another.
Suppose the most powerful AI remains concentrated among wealthy corporations and affluent users. What happens to everyone else? One possibility is dependence. People simply use cheaper and less capable versions of the same technology. But there is another possibility. They may be forced to create their own technological worlds.
There is a possibility that AI systems could develop around particular languages and social needs. For instance, one may see a Hindi AI, a Tamil AI or a Bengali AI built around regional archives and histories emerging in near future. There could be an agricultural AI designed for farmers.
These systems would not necessarily be inferior copies of Silicon Valley’s frontier models. They could be different systems performing different functions. It may mean creating different kinds of intelligence for different social worlds developed locally. Vernacular technology can develop its own purposes, institutions, users and forms of authority. A rural community may need a different AI from a multinational corporation. The future may, therefore, not be about one universal intelligence. It may be about a plurality of intelligences.
Conclusion
Historical analogies can help us ask different questions. Sanskrit did not remain the only language of literary and political expression. Persian did not remain the only cosmopolitan language across the Persianate world. Vernacular literary cultures created new centres of intellectual and cultural production. The internet did not simply reproduce the old library. It created new forms of participation, publication and knowledge circulation. AI may do the same.
The technology that begins by concentrating unprecedented computational intelligence in the hands of a relatively small number of corporations and affluent users may eventually stimulate the creation of alternative, open and vernacular technological ecosystems. Perhaps the future will not be about everyone eventually subscribing to the same AI. It could be about the emergence of many AIs for many intellectual worlds.
The history of knowledge reminds us that centres of authority are rarely permanent. The AI cosmopolis as of now belong to those big giants who control the data centre. But beyond that centre, something else may emerge in the form of vernacular AI in near future, which could be locally rooted, multilingual, multifunctional and created for particular communities and purposes.
————————————————————————-
References:
- Pollock, Sheldon. The Language of the Gods in the World of Men: Sanskrit, Culture, and Power in Premodern India. Berkeley: University of California Press, 2006.
- Pollock, Sheldon. “The Cosmopolitan Vernacular.” The Journal of Asian Studies 57, no. 1 (1998): 6–37.
- Orsini, Francesca. “How to Do Multilingual Literary History? Lessons from Fifteenth- and Sixteenth-Century North India.” Indian Economic and Social History Review 49, no. 2 (2012): 225–246.
- UNESCO. “AI Literacy and the New Digital Divide: A Global Call for Action.” UNESCO, 2024/2025.
- Google. “Introducing Google AI Ultra: The Best of Google AI in One Subscription.” Google Blog, 20 May 2025.
- Singh, Kamlesh. “iPhone Duo: Manuvadi Apple agrees for a cost census.” India Today, 10 September 2026.
Speaking History 
Leave a comment